Contra Zvi on gain of function research

I have some things to say in response to Zvi's latest covid update. I harbor deep suspicions about the suggestion to ban, shun, or otherwise suppress gain of function research. A world in which that advice was taken to heart would be likely be significantly worse than the world we live in.

First, some background. Early in the pandemic, Zvi streamed an enlightening debate with Robin Hanson about whether or not variolation should be used to fight COVID-19. It turned out that neither of them really doubted that variolation as Robin proposed it —healthy people choosing to be infected with low doses of the virus in controlled and isolated conditions in order to build immunity—would likely have been a very effective tool against COVID-19 at that point, and could have saved many lives on net.

And yet, nevertheless, Zvi argued that this was not a worthwhile avenue to pursue. I started out being very much on team Hanson—if a policy of widespread variolation would undoubtedly many save lives, how could I possibly be anything but strongly for it? And yet despite this initial conviction, Zvi's argument completely persuaded me.

The core of the point that convinced me was that you can't blindly talk about a policy in isolation, without talking about what would need to be true for it to be implemented. Zvi rightly argued was that there's no point in focusing on variolation, if only because any society that was sane enough to seriously consider it would be sane enough to do other, even more effective interventions.

In retrospect, it's very clear what a even a moderately sane civilization would have done to beat the pandemic—human challenge trials on the candidate vaccines, and then approve and start distributing them within a few weeks. By March or April 2020 vaccinations could have started, and by then there wouldn't remain much point in even talking about variolation. That doesn't mean that the pandemic would have instantly been over in this scenario—tricky issues related to the logistics of manufacturing and distributing the vaccine at scale would remain—but it would have moved past the stage where discussing variolation would have been more than a distraction.

That's what would have happened in a society sane enough to think things through. A society not sane enough, like ours, wouldn't manage to do either that or variolation—there's no middle ground where variolation is a serious policy proposal.

I extract a general principle from this: don't consider policy proposals in a vacuum, but rather consider the attitude which society would need to have in order to implement that policy, and then figure out what that attitude would imply.

What happens if we take this principle seriously? My claim is that it implies that Zvi's arguments about gain of function research in his latest covid update are misguided at best. Zvi's frames it as a policy suggestion, but the social attitude which that policy would require scares me deeply. I strongly suspect that a world in which it was incrementally easier to stop disfavored types of research would be really bad. Specifically, I think we wouldn't have mRNA vaccines, our main tool in the fight against COVID-19.

We all know by now the story of how Karikó Katalin was demoted and effectively forced out of academia because her disfavored research on mRNA. She barely found a refuge from which she could continue her research, and humanity should count itself extremely lucky for it: if it weren't for the fact that she managed to hang on and find a refuge in industry, we may well not have had the mRNA vaccine against covid.

It's not that Zvi is proposing a ban on mRNA research, of course. But just like an actual implementation of the variolation proposal would have implied a saner world in which variolation itself would have been beside the point, the suppression of GoF would imply a world all the more eager to suppress research. mRNA research only barely escaped being crushed as it is, so you can imagine my doubts about a world which leans further (even slightly!) in the direction of suppressing disfavored research.

One obvious response which I don't think holds much water is that we only should be in favor of suppressing "bad" research like GoF, and against suppressing "good" research like mRNA. This doesn't and can't work—we have access only to the map, not the underlying territory. Policies can't be conditioned on territory-based measures of "good" or "bad", only on the map-based projections of those measures—"favored" or "disfavored". Maybe it could be done if Zvi was Science Czar, but Zvi himself explains why that wouldn't and couldn't work.

Only one worthwhile question remains: given that we don't and can't have Zvi as Science Czar, would I prefer, on net, to be in a world where it was easier to suppress disfavored research? I think that the answer here is a clear "no". It's certainly possible to disagree with me on this last point, but this is at the very least the correct intellectual battleground on which to hold the debate.